The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill functions as a main entry point and router that processes user-provided data (e.g., article drafts, feedback, and conversation history) to determine which sub-skill to execute. This creates a surface where malicious instructions embedded in user data could influence the agent's routing decisions.
Ingestion points: The skill is instructed to "先完整读取本次对话，提取已说过的目标、材料、约束和已完成的步骤" (read the entire conversation, extracting goals, materials, constraints, and steps) to determine the correct mode and route as defined in SKILL.md.
Boundary markers: The skill lacks explicit boundary markers or "ignore instructions" warnings to separate user-provided content from its own operational logic when passing data to sub-skills.
Capability inventory: The router directs data to downstream tools with significant capabilities, such as dby-publish (which writes to the user's WeChat Official Account backend) and dby-api (which performs external network operations).
Sanitization: No sanitization or validation logic is present to filter or escape instructions embedded within user-controlled data before it is interpolated into the workflow.

dby